R

• An open source language developed for statistical computing by Ross Ihaka and Robert Gentleman
• Inspired by the S language developed at Bell labs in 1976 to make interactive data analysis easier
• The first official version was released in 2000

Why use R for data analysis?

• R is and will always be free
• Can easily implement any statistical analysis
• Code serves as a record which enables reproducibility with minimal effort
• As of March 2023, there were over 19,000 open source packages to extend its functionality
  • Highly customizable graphics (ggplot2)
  • Analysis reports (knitr)
  • RNA-seq analysis (DESeq2)

How does it work?

RStudio

• RStudio is an integrated development environment (IDE)
  
• It makes R code easier to write by providing a feature rich graphical user interface (GUI)

Layout

File types

• Rscript files that end in .R
  • The most basic, a file that contains R code
• RMarkdown files that end in .Rmd
• Let’s create a blank Rscript to see how they work, open RStudio and click:
  • File -> New File -> R Script

R Markdown

• A file format combining R code with Markdown for text formatting.
• Designed for creating reproducible research reports in various formats (HTML, PDF, Word).
• Let’s create an Rmd file in RStudio to explore the basics of how they work:
• File -> New File -> R Markdown

R Markdown Advanced Usage

• Presentations: Creating slides (like these) with revealjs.
• Publications: Authoring online books that combine narrative, code, and output with bookdown.
• Interactive Documents: Developing interactive tutorials or dashboards with learnR and other embedded applications.

Variable definition

• Variables store information that is referenced and manipulated in a computer program
• In contrast to the mathematical definition of a variable, variables in computer science are mutable
• There are 3 ways to define variables in R, but one is preferred:

x <- 1  # Preferred way
x = 1
1 -> x
print(x)

## [1] 1

Variable naming

• Variables names must start with a letter and can contain underscores and periods
• It is best practice to use descriptive variable names and stick to one style of names

# Snake case
dog_breeds <- c("Labrador Retriever","Akita", "Bulldog")

# Period separated
dog.breeds <- c("Labrador Retriever","Akita", "Bulldog")

# Camel case
DogBreeds <- c("Labrador Retriever","Akita", "Bulldog")

Poll 2

Which variable name is not valid in R?

1. cat_dog
2. CatDOG
3. cat.dog
4. catD0g

Excercise 1

• Open Rscript file part_1.R in Rstudio

Data Types

• Integer
  • Whole numbers
• Numeric
  • Decimal numbers
• Logical
  • Boolean (TRUE, FALSE)
  • NA
• Character
  • Letters and strings of letters
  • “A”, “Labrador Retriever”

Data Structures

• Vectors
  • Atomic vectors - one dimensional lists that store values of the same type
  • Lists - can be multidimensional and contain different types/structures (ex. nested lists)
• Factors
  • Ordered list with assigned levels
• Matrix
  • Columns and rows of the same type
• Data frames
  • Columns and rows of mixed types

Exercise 2: Data Types and Structures

• Reopen Rscript file part_1.R in Rstudio

Math & Logic

• Built in functions to get common mathematical summaries of data (eg. mean( ), median( ), mode( ) )
• Relational comparison operators to compare values

x == y  # Equal to
x != y  # Not equal to
x <  y  # Less than
x > y   # Greater than
x <= y  # Less than or equal to
x >= y  # Greater than or equal to

x %in% y # Is x in this vector y?

Logical Operators

• Logical operators can compare TRUE or FALSE values

x <- TRUE
y <- FALSE

!x     # Not x
x | y  # x or y
x & y  # x and y

Conditional execution

• Relational and logical operations allow for conditional execution of code

if ("Akita" %in% dog_breeds) {
  print("dog_breeds already contains Akita")
} else {
  dog_breeds <- c("Akita", dog_breeds)
}

## [1] "dog_breeds already contains Akita"

Functions

• A function is block of organized, reusable code that is used to perform a single action
• R has many built in functions, these are called base R functions
• Not all arguments are required and some have default values

Defining a function

• To define a function we use the function keyword, the output is specified with the return keyword:

add_dog <- function(dog_to_add,
                    input_vector) {
  if (dog_to_add %in% input_vector) {
    
    message("Already contains this dog")
    
  } else {
    
    output <- c(dog_to_add, input_vector)
    return(output)
    
  }
}

Example

add_dog(dog_to_add = "Akita",
        input_vector = dog_breeds)

## Already contains this dog

add_dog(dog_to_add = "German Shepard",
        input_vector = dog_breeds)

## [1] "German Shepard"     "Labrador Retriever" "Akita"             
## [4] "Bulldog"

Packages

• Packages are collections of functions that are specialized to a specific task (plotting, data manipulation etc.)
• The tidyverse is a collection of commonly used data analysis packages - Learning curve is less steep - Lots of useful packages for data analysis

Workshop survey

• Please fill out our workshop survey so we can continue to improve these workshops

Upcoming Workshops

1. Introduction to Statistics, Experimental Design, and Hypothesis Testing
   • Jan 25, 2024 (Session 1 - 10am–12pm) (Session 2 - 1pm–3pm)
   • Jan 26, 2024 (Session 3 - 10am–12pm)
2. Intermediate RNA-Seq Analysis Using R
   • Feb 1, 2024 (9:30am-12:00pm)

General Tips

• Always confirm ChatGPT’s outputs are correct
• Provide as much detail as possible about the problem in the 1st prompt
• Use separate chats for separate tasks/projects
• Try the ‘Custom Instructions’ function that adds additional information to every prompt
• Can visit webpages (GPT 4 only), which can help get more specific answers

Code Tips

• Commented R code yields better responses in my experience
• Provide the code and error message in the same prompt
• ChatGPT can work well to convert syntax and improve your code:
  • “Turn this loop into a function : [your code]”
  • “Is there a better way to do this : [your code]”
• Check out the file: example_code/1_convert_syntax_example.R for an example use case

Key Questions

• What assay was the package designed for?
• When was the last release?
• Is it maintained (frequent updates)?
• Does it work on all operating systems?
• Are other people using it? (citations)
• Do they respond to github issues?
• Is there a benchmarking paper?

BioConductor and CRAN

• Both of these have stringent requirements for packages they host (eg. for BioConductor they have to run on all major operating systems)

• Prefer BioConductor packages if available over CRAN

• Prefer CRAN packages over ones only hosted on GitHub

Start with the Assay

• Click here to go to BioC views
• Pick the assay you want to analyse
• Pick the type of analysis you want to do
• Find a package that does it
• Find benchmarking papers to narrow the list of packages down
• Find the vignette on the package page and refer to the manual for any questions not covered by it

R

• R Markdown: The Definitive Guide : Excellent R markdown reference

• R for Data Science

• ggplot2: elegant graphics for data analysis

• Advanced R

Statistics

• Data Analysis in R : This book has more statistics details than R for Data Science
• Generalized Linear Models
  
• Random Effects

RNA-seq Analysis

• RNA-seqlopedia : Everything you need to know about RNA-seq experiments
• RNA-seq Expression Units : Blog post on understanding common units
• Introduction to Single-Cell Analysis with Bioconductor : Covers the basics of scRNA-seq analysis in R

Dimensional Reduction

• Tutorial on PCA : PCA explained with R code examples
• Understanding UMAP : Short explanation with great visualizations, mainly useful for scRNA-seq analysis